Contribution au diagnostic des défauts de la machine asynchrone

GUEDIDI, ASMA (2024) Contribution au diagnostic des défauts de la machine asynchrone. Doctoral thesis, Faculté des Sciences et de la technologie.

Text THESE_finale.pdf Download (13MB)

Abstract

Induction motors (IMs) are prevalent in both industrial and domestic applications. The implementation of an automatic condition monitoring system for IMs proves invaluable in detecting faults at their initial stages. This proactive strategy aims to prevent machinery malfunctions and potential catastrophic failures. Despite decades of research employing various approaches for IM fault diagnosis, the accurate identification of faults remains a complex task due to the intricate signal transmission paths and the influence of environmental noise. The primary objective of this thesis is to create an innovative intelligent system that enhances the reliability of health condition monitoring for induction motors (IMs) considering two cases: 1) The first one aims to design an automated fault diagnosis system utilizing ANN models to confront the complication of overlapping data in identifying broken rotor bar faults. Diverse signal processing techniques are incorporated into the system to refine and optimize the diagnostic accuracy for more effective fault detection. 2) The second one aims to create an automated fault diagnosis based on conventional neural network models. In this part, we focused on improving the performance of diagnostic systems. To this end, two solutions were proposed. Firstly, we aim to reinforce the quality of the images by merging the data images with the information map. Secondly, we focused on the architecture of an improved SqueezNet model associated with an attention block, which gives high precision in image classification. This model was validated by data relating to short-circuit and eccentricity faults. The results obtained are surprising

Actions (login required)

View Item